In the manufacture of integrated circuits, and particularly in single wafer manufacturing, one of the most important practical technique considerations is the throughput; i.e. the speed with which devices can be manufactured, or considered from another standpoint, the number of devices that can be produced in a given amount of time. Processes that require long processing times are less practical from a manufacturing standpoint and processes or techniques which reduce process times while retaining or increasing the quality of resulting devices are almost always advantageous.
In the single wafer processes for which throughput is important, one of the most common thin films used in integrated circuit manufacture is polycrystalline silicon, often referred to as "polysilicon" or "poly-Si." For example, in metal-oxide-semiconductor (MOS) integrated circuit technology, polysilicon doped to be sufficiently conductive is typically used as the gate electrode and as an interconnect material on silicon dioxide. It thus offers a wide range of practical applications for which it has been found most useful.
Nevertheless, there are disadvantages to the use of polysilicon in single wafer manufacturing, the most glaring of which is the relatively low deposition rate of silicon at the temperatures at which the best material properties are obtained. Typically, this means that the deposition of polysilicon requires a high "thermal budget," a term used to describe the combined product of time and temperature required to carry out any particular process. For example, polysilicon typically must be deposited at temperatures greater than about 550.degree. C. (usually about 575.degree. to 650.degree. C.) in order to exhibit acceptable deposition rates. Even so, these deposition rates are very slow, and can only be increased by raising the deposition temperature. The use of higher temperatures, however, causes other problems in the manufacturing process. In some instances exposure to higher temperatures requires that other materials in the integrated circuit that are sensitive to such temperatures be protected or eliminated from the effects of such a heating step. As another illustration, dopants present in the semiconductor material will exhibit an increased tendency to diffuse or migrate as higher temperatures are used, thereby destroying the p-n junction characteristics which make the device operable in the first place. Furthermore, a number of metals cannot be present during a processing step as they would melt and reflow at such temperatures.
Other disadvantages of the silicon processes include the typical characteristic that in a single wafer processor using an optical heating process ("optical furnace") and the appropriate quartz windows, polysilicon will deposit on these hot windows as well as on the wafers in the optical furnace. The windows then become heated and in turn block the optical heating of the wafers, thus causing additional secondary problems in the overall deposition process. For example, temperature measurement using optical pyrometers tend to become erroneous under such circumstances.
Another disadvantage of polysilicon is that a preferred manufacturing process uses low pressure chemical vapor deposition (CVD) which slows the deposition of polysilicon even further. As is known to those familiar with this technology, CVD can take place under various conditions of temperature and pressure. At higher pressures, however, the silicon carrying gas (usually silane, SiH.sub.4) tends to deplete from wafer to wafer in the commonly used multi-wafer techniques. Atmospheric pressure CVD, or other higher pressure CVD processes, also tend to produce wafers which exhibit the depletion of silane even across single wafers. In single wafers, the depletion effect is referred to as a "bulls eye" problem. Low pressure CVD tends to minimize or eliminate these problems, but slows the deposition rate as just described.
All of these disadvantages, among others, provide difficulty in the use of polysilicon in certain device manufacturing techniques with accompanying compromises in device design or manufacturing through-put.
One potential alternative to the use of polysilicon, particularly in MOS devices, is the use of germanium (Ge) as a substitute for silicon in certain applications. Germanium has electrical properties generally similar to silicon and offers several theoretical advantages as well, particularly when analyzed in terms of a thermal budget. For example, when using low pressure CVD, germanium can be deposited at temperatures as low as 350.degree. C. at considerably higher deposition rates than can silicon at much higher temperatures. Thus, a layer of polycrystalline germanium of a given thickness can be deposited more quickly and at a lower temperature than can an equivalently thick layer of polysilicon. The lower thermal budget for germanium provides resulting advantages for through-put in device manufacturing processes.
Using present technology, it is recognized that thin films of germanium can be selectively deposited on silicon using low temperature chemical vapor deposition of germane (GeH.sub.4) as the reactive gas. Under typical low pressure CVD conditions, however, germanium will not deposit on silicon dioxide (SiO.sub.2) which is typically used as a localization barrier in the manufacture of integrated circuit devices and therefore often adjoins a silicon surface. The tendency of germanium to refrain from depositing on silicon dioxide is apparently due to the low density of germanium absorption sites available on SiO.sub.2. Additionally, it may be possible that this characteristic results from a GeH.sub.4 etching of the SiO.sub.2 surface that produces volatile GeO.
As a result, to date the use of germanium in certain kinds of structures has been limited by the characteristic that germanium will not deposit on silicon dioxide in the otherwise convenient processes of chemical vapor deposition using germane gas. Thus, in metal-oxide-semiconductor (MOS) structures such as depletion and inversion mode MOSFETs or MOS capacitors, the gate electrode material has been limited to polysilicon or some other material rather than the thermally advantageous polycrystalline germanium material.
The need therefore exists for a method of advantageously using the properties of polycrystalline germanium, particularly the low thermal budget with which it can be worked, in combination with MOS-type structures.